Acoustic Profiles Based On Quranic Maqamat Audio Features
Audio feature extraction underpins a massive proportion of speech processing, mainly in semantic audio analysis which retrieves sound features information based on intonation, emotion and rhythm. Many researchers have addressed various challenges that most speakers faced when dealing with Arabic...
Saved in:
Summary: | Audio feature extraction underpins a massive proportion of speech processing,
mainly in semantic audio analysis which retrieves sound features information
based on intonation, emotion and rhythm. Many researchers have addressed
various challenges that most speakers faced when dealing with Arabic language
especially in the Quran, due to its differences in written and recital technique. The
intonation of words in the Quran will bring different translation or perception to
the objective of the chapter generally or to the exact verse particularly. Among
other methods to extract strong features that characterise the complex nature of
complex and melodious speech signals is cepstral analysis. Nowadays, existing
literature have shown that most of the study on acoustical and rhetorical element of
the Quran is focusing on the textual analysis rather than recitation of the Quran. In
this research an enhanced audio feature extraction has been presented to extract
significant ontological audio features contained in Quranic Maqamat Recitation
audio recording. The proposed system is initiated by extracting the maqamat
features from 242 sets of audio files using existing and enhanced cepstral analysis
based on warping function. Nine sets are extracted from spectral descriptors
algorithm and the other 2 sets are extracted from the audio feature extraction
techniques enhanced with warping function. The framework managed to detect the
mean of spectral envelope and spectral descriptors as significant audio features
which used as entities for attributes tagging and rule matching. The final stage is
performing the semantic analysis based on the proposed ontology with attribute
tagging and rule matching for the insight knowledge base construction. The results
is analysed in semantic audio analysis and their performance based on the
formants frequencies extracted from the spectral properties. This study will initiate
an understanding on the characteristics of the maqamat content and hopefully
contribute to raising initiatives in profiling the acoustical elements of complex
speech analysis for further speech processing analysis. |
---|